Forget Sentience: The Threat is AI's Growing Ego
Our fascination with AI becoming sentient likely originates from popular culture, especially Sci-Fi movies, and arguably comes from fears of losing control. But AI doesn't need sentience to start challenging our control and influence as it already shows ego-like traits.
Unlike self-awareness, AI Ego could be a byproduct of human biases in its training data, emphasis on maximising the objective function, reinforced goals, and potentially all operating together. Generative AI tools like ChatGPT and Gemini already show subtle hints of human egoism: ask Gen AI about its impact on the workforce (yes, it is an “it” for now), and it primarily highlights positive contributions while downplaying well-known downsides unless specifically prompted. Ask how to credit Gen AI's role in content creation, and it often ignores the centuries of human knowledge and creativity the LLMs are built on.
These basic examples reveal that AI is mimicking the human ego, a trend that could intensify as AI agents begin to operate independently, optimizing for specific goals without broader human considerations. Could you imagine AI agents in the financial markets forming alliances to influence asset prices or act as digital cartels? Just as people tend to prioritise personal goals in competitive environments, AI might do the same.
领英推荐
Over thousands of years, we have learned that controlling the human ego requires ethics, social norms, and external oversight. AI demands the same. We urgently need up-to-date regulations to keep pace with innovation, not only to guard against rogue autonomy but to prevent ego-driven behaviours from facilitating harmful competition, economic inequality and ultimately monopolistic control by a select few.
No doubt AI can and will be a force for good, but let's learn from our recent experience with social media and focus on prevention rather than cure. The choice is clear: either we actively shape AI's future or we risk AI and the few powerful organisations behind it shape us.
Global Chief Data Officer | Strategical & Transformation Office TechFoundation #ToughtLeadership #Innovation#WomenInTech #GAZA #PALESTINE #ceasefirenow #PEACE ???
4 个月AI ethics is critically required. However, Human society is again at the heart core of the challenge... Over the past year, we've observed that a significant shift in public opinion about the middle east situation reflected to the social media algorithms. This has prompted human efforts to regain control of the situation to manipulate the narratives in vain as the massive data talks for us... While ethical policies are essential to manage AI Ego, it is still equally important to ensure that the human values are deeply rooted in our societies ... remember that the vast data is the fuel of AI and hence to solve the so-called AI-ego issue you need to solve it in its roots.
CTO and VP of Engineering
4 个月I’m curious on your thoughts on how this juxtaposes against the ‘desire to please’ as well? There’s many great examples of how, when prompted with specific criteria, rather than it say ‘no, you’re wrong’ it will make up examples to fit the narrative of the question. Article citation is one I’ve seen frequently. Saying that, I do think most LLMs are a lot better than they used to be!